Abstract:

The presented work provides an analysis of various regression models constructed by methods of neural networks in typological groups. As observations, the regions of the Russian Federation are considered, divided into three clusters, depending on the level of their economic development. The exogenous variables are factors such as: an integral characteristic of tension in the residential real estate market, an integral characteristic of the regional economy, affecting the housing market and housing construction, as well as an integral characteristic of the comfort of living quarters. The endogenous variable is such an indicator as the volume of commissioning of residential buildings. To determine the best neural network regression models, the following types of neural network models were considered in the work: linear network, generalized regression network, network based on radial basis function and multilayer perceptron.